The CNN Hip Accelerometer Posture (CHAP) Method for Classifying Sitting Patterns from Hip Accelerometers: A...

Area of Science:

• Gerontology
• Biomedical Engineering
• Wearable Technology

Background:

• Sitting patterns are crucial for healthy aging outcomes.
• Hip-worn accelerometers can measure these patterns, but current methods struggle with postural transitions.
• Accurate measurement of sitting behavior is essential for understanding its impact on aging.

Purpose of the Study:

• To develop a novel method for accurately classifying sitting patterns and postural transitions using hip-worn accelerometers.
• To overcome the limitations of existing methods in detecting sit-to-stand transitions.
• To enhance the analysis of older adult movement data for healthy aging research.

Main Methods:

• Developed the Convolutional Neural Network Hip Accelerometer Posture (CHAP) classification method.
• Utilized data from 709 older adults wearing hip-worn ActiGraph GT3X+ accelerometers and thigh-worn activPAL inclinometers.
• Compared CHAP against traditional cut-point methods and a previous two-level behavior classification (TLBC) algorithm.

Main Results:

• CHAP achieved 93% agreement for minute-level sitting classification, outperforming other methods (74%-83%).
• CHAP demonstrated superior sensitivity (83%) and positive predictive value (83%) for detecting sit-to-stand transitions compared to cut-point (73%, 30%) and TLBC (26%, 71%).
• Day-level sitting pattern metrics derived from CHAP closely matched activPAL data, unlike cut-point and TLBC methods.

Conclusions:

• CHAP is the most accurate method for classifying sit-to-stand transitions and sitting patterns from hip-worn accelerometer data in older adults.
• This advancement enables more precise measurement of sitting patterns, facilitating large-scale studies on healthy aging.
• CHAP promotes enhanced analysis of older adult movement data, contributing to a better understanding of sedentary behavior's impact on aging.